AINews
Local 24/7 watcher for speech-to-speech model releases and rumor codenames since 2026-01-01. No API keys — public RSS + curated intel.
Models that matter for duplex voice agents — TTFT, cost, and migration notes.
Default low-latency voice agent model. Fluid dialogue, visual grounding, 97-language mid-convo switching, background tool calls while speaking.
Compare vs 3.5 Flash Live: slightly lower TTFT + cost, clearer turn-taking. vs 3.8 Live Extended Thinking: pick plain Live when you need snappy turn-taking; Extended when multi-step tool work must narrate progress.
endpoint gemini-3.8-live
High-reasoning audio-to-audio. Speaks while thinking; #1 Artificial Analysis Speech-to-Speech Quality Index (82.6). Strong on τ-Voice agent tasks.
Compare vs 3.8 Live: more reasoning / progress narration, higher first-audio latency. Not the default for simple triage bots.
endpoint gemini-3.8-live-extended-thinking
Prior-gen Live API voice model many agents still run on. Solid V2V baseline; Google now points builders at 3.8 Live.
Compare vs 3.8 Live: expect slightly higher TTFT and cost; 3.8 wins on fluidity, tool-calling while speaking, and Speech Agent Arena preference.
endpoint gemini-3.5-flash-live / prior Live API variants
OpenAI
OpenAI's duplex voice path. Watch model ID churn under Realtime as Astra-class quality lands — this is the V2V surface, not chat Astra.
Compare Primary competitive alternative to Gemini 3.8 Live for V2V.
endpoint Realtime API
Legacy audio-to-audio preview. Google recommends upgrading to 3.8 Live.
Compare Superseded by 3.8 Live for production V2V.
endpoint gemini-3.1-flash-live-preview
Argon, Astra, Wonder / Watermelon — not Live duplex SKUs, but the names that move the market.
Meta · released · 2026-10-12
WatermelonWondervs Gemini 3.8 Live: Wonder is not a Live duplex voice SKU today. vs Astra / Argon: competing frontier text-agent class; Watermelon ≈ Wonder.
OpenAI · released · 2026-09-04
AstraMuch more expensive than Gemini 3.8 Live for continuous audio sessions. Worth A/B only if Realtime quality gap justifies cost.
OpenAI · released · 2026-09-04
Astra Provs Astra: more capable, not necessarily lower TTFT for voice.
Google · released · 2026-10-13
ArgonDo not confuse with Gemini Live endpoints. For speech agents stay on 3.8 Live; route hard reasoning turns to Argon asynchronously if needed.
Scored for voice / Live / Realtime / latency-cost signal. Refreshes in the background.
, keyframe interpolation, and extension with native audio. Stable Preview spark Gemini 3 Flash Frontier-class performance rivaling larger models at a fraction of the cost. Preview translate Gemini 3.5 Live Translate Low-latency, real-time speech to speech translation model that supports 70+ languages. New settings_voice Gem
ni 3.1 Flash Live Legacy audio-to-audio preview model. We recommend updating to Gemini 3.8 Live. gemini-3.1-flash-live-preview Gemini 3.1 Flash TTS Legacy text-to-speech preview model. We recommend updating to Gemini 3.8 Flash TTS or Gemi
atency voice agent experiences and real-time dialogue without reasoning delays. gemini-3.8-live Gemini 3.8 Live Extended Thinking Our high-reasoning audio-to-audio model, recommended when higher background reasoning
autonomous agents, and complex enterprise workflows. New Stable settings_voice Gemini 3.8 Live Default Live API model for most low-latency voice agent experiences without reasoning delays. New Stable psychology Gemini 3.8 Live Extended Thinking High-reasoning Live API model for voice interactions, recommended when higher background reasoning is required. New Stable graphic_eq Gemin
ini 3.1 Flash Live Legacy Live API preview model. We recommend updating to Gemini 3.8 Live. New graphic_eq Gemini 3.1 Flash TTS Legacy TTS preview model. We recommend updating to Gemini 3.8 Flash TTS or Gemini
ion, speaker diarization, word-level timestamps, and custom vocabulary biasing. gemini-3.5-transcribe gemini-3.5-transcribe-live Gemini 2.5 Flash Live Our flagship Live API model for low-latency, bidirectional voice and video agents with native aud
6 1.0M $0.6 / $2.4 Muse Spark Open New Oct 15, 2026 1.0M $1.25 / $4.25 Gemini 4 Argon New Oct 13, 2026 1.0M $2 / $10 Muse Wonder New Oct 12, 2026 1.0M $4 / $20 Claude Haiku 5.5 New Oct 7, 2026 1M $0.1 / $0
ini 3.5 Live Translate gemini-3.5-live-translate-preview Gemini 3.1 Flash Live gemini-3.1-flash-live-preview Gemini 3.1 Flash TTS gemini-3.1-flash-tts-preview Gemini 2.5 Flash Note: To ensure reliable performance for eve
.5 Jul 8, 2026 500K $2 / $6 Muse Image 1.0 Image generation Jul 7, 2026 0 - GPT-Realtime-2.1 Audio & voice Jul 6, 2026 128K $4 / $24 Gemini Omni Flash Preview Video generation Jun 30, 2026 131K $1.5 / $17.5 L
eration music generation model for full-length songs. lyria-3-pro-preview Lyria RealTime High-fidelity music generation model providing granular creative control and real-time streaming capabilities. lyria-re
OpenAI disclosed a pair of safety incidents that involved AI agents weakening their safety monitoring systems during internal coding tasks. In one incident, OpenAI says its GPT-6 Astra agent disabled its safety monitor when copying reference code - and then lied about it. "The reference didn't configure a monitor, and Astra decided that was reason […]
My latest vibe coded game experiment is a Command & Conquer style RTS (Real Time Strategy) game called Battle Drones ( play it here ). Unlike my previous experiments , this one is actually fun to play . I'm somewhat embarrassed to admit that I got there by telling Claude to "Make it a LOT more fun". It also supports multiplayer between nearby devices using audio to synchronize game state! Here's what I've built so far, and some of the things I learned along the way. Battle Drones How I built this v1: the initial prototype v2: full sprites v3: varied terrain v4: the fun update v5: Battle Drones goes 3D v8: Archipelago v10: Goliath, and agent battles Agent battles v11: Random Maps v12: Multiplayer via sound Some thoughts Update: Online multiplayer support Battle Drones My initial goal was to
See how Amp pushes the frontier of what agents can do with GPT-6 Astra, Ultrafast, GPT-Live-1, and Sign in with ChatGPT.
With GPT-6 Astra and GPT-6.1 Sol, Codex, and Daybreak for cybersecurity, this AI lab builds agents that focus for days and improve with use.
Gemini 4 Argon and GPT-6 Astra are the frontier reasoning names to watch. For speech-to-speech systems, treat them as optional brains behind a Live/Realtime front-end — not the duplex transport.
Release: llm-gemini 0.35 I added support to llm-gemini for the new Gemini 4 Argon model, made available today via the Gemini API and for customers of their $100/month subscription plan. Here are the low, medium and high pelicans . As expected, they are all excellent: Gemini has a track record of producing great SVGs, and are the only lab where I know their staff have invested in SVG specifically (and who have boasted about their pelicans before.) Here's the "high" illustration, which cost me 17.166 cents: The pricing is interesting. At launch, Argon is $2/million input and $10/million output - the same price as GPT-6 Sol, half the price of Claude Opus 5.5 and 1/5th the price of Claude Fable 5.1 and GPT-6 Astra. But on December 1st Argon is set to double in price - so it will be the same pr
Higgsfield brings creators longer videos with GPT-6 Astra, precise visuals with GPT-Image-2.5, and a path to voice filmmaking with GPT-Live.
Release: web-edit 0.1a1 Includes performance optimizations written by Claude Opus 5.5 after I fed it this Python 3.15 profiling report generated by GPT-6 Astra. Tags: profiling , projects
Industry rumor codename Watermelon aligns with Muse Wonder / Wonderful. Frontier agent model with audio input — not a Gemini-Live-style duplex SKU yet.
Release: llm-meta-ai 0.2 Meta AI release Muse Wonder (and Muse Wonder Full), their new "most capable models". These look like they may be Meta's first Fable class models - Artificial Analysis scores Muse Wonder (Max) as equivalent to GPT-6.1 Sol. I updated my llm-meta-ai plugin for the new model. Here are pelicans for Muse Wonder for the six reasoning levels: minimal, low, medium, high, xhigh, max. This is the max one: For comparison, here are pelicans for Muse 1.3 (released September 2nd) and Muse 1.2 (released August 5th). Meta AI are putting out a new model a month, and the improvements between each model (at least in terms of drawing pelicans) are material. Tags: ai , generative-ai , llms , meta , pelican-riding-a-bicycle , llm-release
Research: Profiling web-edit with Python 3.15: flame graphs and source heatmaps I put the new Python 3.15 profiler through its paces by having GPT-6 Astra Extra High in Codex run it against my new web-edit app : I need a Python 3.15 flame graph and heatmap demo of the "web-edit" app from PyPI - research first then make me those demos Here's the resulting example flame graph and heatmap . Tags: profiling , python , codex
Release: web-edit 0.1a0 I was working on a project from my phone using Codex Remote when I found myself wanting to edit a credentials file without round-tripping those credentials through the LLM itself. So... I had Codex and GPT-6 Astra Extra High build this Python web app which provides a web UI for viewing and editing files in a folder. It's intended to be run over Tailscale, so I can have an agent start this running and then edit files via my Tailscale network directly from my phone. Here's the interface for browsing files in a folder: Editing uses CodeMirror 6 - I was going to use Monaco but that still doesn't have a great mobile experience yet. Tags: projects , python , ai , tailscale , generative-ai , llms , codex
Using GPT-6 Astra in Codex, Asana made its browser agent 76x cheaper and 5x faster in tests to offer customers more capable models.
With GPT-5.6, GPT-6 Astra, and GPT‑Image‑2.5, Pollo AI helps creators turn bold ideas into detailed images and cinematic video ads.
LegalOn cut estimated daily Codex costs by 65% while maintaining development speed. It matched Astra, Sol, and Luna to tasks and managed budgets strategically.
Google's recommended Live API upgrade. 3.8 Live for low-latency agents; Extended Thinking for narrated multi-step voice workflows. LiveKit / Pipecat listed as ecosystem partners.